# Original Paper vs. Followup: Full Comparison (Post-IFS Correction)

## Setup

| Dimension | Original Paper | Followup (Corrected) |
|:----------|:---------------|:-----------------|
| Countries (demographics only) | 69 (EBA-49 + 20 SSA) | 141 (+10 EU, +62 expansion) |
| Countries (baseline model) | 67 | 137 |
| Countries (extended model) | 44 | 90 |
| Observations (baseline) | 1,857 | 2,695 |
| Observations (extended) | 961 | 1,592 |
| Population coverage | 93.3% | 96.9% |
| GDP coverage | 98.6% | 99.1% |
| IFS-to-ISO3 mapping | 57/148 codes wrong (39%) | **Corrected** |
| Methodology | Pooled GLS, AR(1) Cochrane-Orcutt | **Identical** |
| Demographic variables | Fair-Dominguez Z₁, Z₂, Z₃ | **Identical** |
| EBA controls | fiscal, KAOPEN, NFA, OPW, health, LE | **Identical** |

## Core Demographic Results

| Finding | 69 (corrupted) | 69 (corrected) | 140 (corrected) | Verdict |
|:--------|:--------------|:---------------|:--------------------------|:--------|
| Z₁ (demo only) | 25.2** (p=0.045) | 15.3 (p=0.284) | 37.1*** (p<0.001) | Corrupted→NS; expansion recovers |
| Z₁ (parsimonious 4 ctrls) | — | 31.9 (p=0.135) | **38.4** (p=0.017) | Expansion needed for power |
| Z₁ (full baseline) | 29.4 (p=0.097) | 15.3 (p=0.483) | 19.0 (p=0.238) | **NS in all corrected specs** |
| Baseline R² | 0.309 | 0.395 | 0.284 | Corrected 69c fits better |
| Key absorber | — | — | log_rel_opw (Z₁: 38.4→14.9 when added) | Bad-controls problem |

## Control Variables

| Variable | Original (69) | Followup (140, corrected) | Change |
|:---------|:--------------|:--------------------------|:-------|
| fiscal_bal_gdp | 0.30 (p<0.001) | 0.31 (p<0.001) | Stable |
| kaopen | 0.06 (p=0.74) | -0.39 (p=0.078) | Marginal, sign change |
| nfa_gdp_lag | 0.18 (p=0.14) | **-0.04 (p=0.875)** | **Collapsed** (IFS correction) |
| log_rel_opw | 1.03 (p=0.094) | **4.23 (p<0.001)** | 4x larger, absorbs Z |
| health_exp_gdp | -0.19 (p<0.001) | **-0.66 (p=0.003)** | 3.5x larger |
| life_expectancy | -0.19 (p<0.001) | -0.01 (p=0.868) | **Collapsed** |

## Z x KAOPEN Interactions

| Finding | Original (69) | Followup (140, corrected) | Verdict |
|:--------|:--------------|:--------------------------|:--------|
| Z₁ x KAOPEN | 38.6 **(p=0.004)** | 21.6 (p=0.051) | Marginal |
| Z₂ x KAOPEN | -6.1 **(p=0.002)** | -3.6 **(p=0.034)** | Weaker but sig |
| Z₃ x KAOPEN | 0.26 **(p=0.001)** | 0.15 **(p=0.026)** | Weaker but sig |
| Joint F-test | p<0.001 | p<0.001 | Robust |
| Extended R² | 0.219 | 0.339 | Higher |
| **Where they operate** | Assumed universal | **Middle income mainly** | NEW |
| High-income Z₁xKxHigh | Not tested | -20.0 (p=0.70) | Zero effect in AEs |

## Pension Interactions (NEW in followup)

| Finding | Original (69) | Followup (140, corrected) |
|:--------|:--------------|:--------------------------|
| Z x pension tested? | On EBA-49 only | Yes, systematically |
| Z₁ x pension (EBA-49) | — | 9.0 (p=0.116) |
| Z₂ x pension (EBA-49) | — | -1.3 (p=0.089) |
| Z₃ x pension (EBA-49) | — | 0.05 (p=0.071) |
| Joint F (EBA-49) | — | **F=14.32, p<0.001** |
| Horse race vs KAOPEN | — | **Pension wins** (KAOPEN all p>0.29) |
| Z x pension (full 140) | — | **All p<0.02** |

## Nonlinearity Tests

| Test | Original (69) | Followup (140, corrected) | Verdict |
|:-----|:--------------|:--------------------------|:--------|
| NFA² quadratic | p=0.038 | **p=0.013** | **Stronger** |
| NFA creditor (>0) | 0.73 **(p=0.0002)** | -0.22 (p=0.47) | **Collapsed** |
| NFA debtor (<0) | -0.14 (p=ns) | 0.44 (p=0.35) | NS in both |
| Life expectancy² | **(p=0.0006)** | 0.001 (p=0.43) | **Collapsed** |

## GE Clearing

| Metric | Original (69) | Followup (140, corrected) |
|:-------|:--------------|:--------------------------|
| PE imbalance at 2040 | +0.47 pp | +0.47 pp |
| PE imbalance at 2050 | +0.07 pp | +0.07 pp |
| Δr* cap binds | 2035-2040 | 2035-2040 |
| Per-country GE adjustment | 0.2-0.4 pp | 0.2-0.4 pp |

## Projections (selected countries, pp of GDP)

| Country | Original 2025 | Followup 2025 | Original 2050 | Followup 2050 |
|:--------|---:|---:|---:|---:|
| CHN | +1.2 | -3.9 | -0.2 | +5.1 |
| IND | -2.1 | -2.5 | +1.0 | -5.6 |
| JPN | -0.8 | +10.5 | -1.0 | +16.5 |
| KOR | +0.5 | -1.7 | -1.0 | +14.6 |
| USA | -0.3 | +1.0 | +0.2 | +3.9 |
| DEU | +0.2 | +3.7 | +0.6 | +10.8 |
| NGA | -3.8 | -0.7 | -0.1 | -4.8 |

## Coefficient Stability

| Finding | Original (69) | Followup (140, corrected) |
|:--------|:--------------|:--------------------------|
| Jackknife tested? | No | Yes, leave-one-region-out |
| Z₁×KAOPEN range | — | [9.5, 28.5] |
| Sensitive to dropping | — | S/SE Asia, SSA (Z₁×K loses sig) |
| CCA effect | — | Non-commodity CCA still drives Z significance in baseline |

## Bottom Line

| | 69 (corrupted) | 69 (corrected) | 140 (corrected) |
|:--|:---------|:---------|:---------|
| Core Z₁ (full baseline) | p=0.097 | **p=0.483** | p=0.238 |
| Core Z₁ (parsimonious) | — | p=0.135 | **p=0.017** |
| NFA persistence | 0.18 (p=0.14) | **-0.44 (p=0.031)** | -0.04 (p=0.88) |
| Z₁×KAOPEN | 38.6 (p=0.004) | 32.0 (p=0.026) | 21.6 (p=0.051) |
| Financial openness | Universal | Survives correction | Developing-country only |
| Pension interactions | Not tested | Not tested | **Joint F p<0.001** |
| Bad controls | Not tested | Not tested | **log_rel_opw absorbs Z signal** |
| Original Z₁=29.4 was... | — | **Partly artifact of corrupted NFA** | — |
| NFA creditor asymmetry | Strong (p=0.0002) | **Collapsed (p=0.47) — IFS correction** |
| LE convexity | Strong (p=0.0006) | **Collapsed (p=0.43)** |
